Abstract: ObjectiveTo investigate the predictive ability of ultrasound radiomics features combined with peripheral blood miRNA-34a expression levels for pathological complete response （pCR） in breast cancer patients receiving neoadjuvant chemotherapy （NAC）. MethodsA retrospective analysis was conducted on 93 breast cancer female patients diagnosed with breast cancer and treated with NAC from January 2019 to December 2023. Among them, 53 patients were from the Sixth People’s Hospital of Shanghai, and 40 patients were from Dongguan Binhaiwan Central Hospital. The expression level of miRNA-34a in peripheral blood was detected using real-time fluorescence quantitative reverse transcription polymerase chain reaction （qRT-PCR）. 107 radiomics features were extracted from preoperative ultrasound images using Pyradiomics software. After Spearman rank correlation test, Z-score normalization and LASSO regression analysis, five key radiomics features were selected. Clinical models based on miRNA-34a, ultrasound radiomics models based on radiomics features, and a combined model integrating both were constructed. The diagnostic performance was evaluated using the K-nearest neighbor （KNN） classifier. ResultsUnivariate analysis showed that the expression levels of miRNA-34a in the pCR group were higher than those in the Non-pCR group（P < 0.001）. Multivariate Logistic analysis revealed that elevated miRNA-34a expression was an independent risk factor for pCR in breast cancer patients receiving NAC （P = 0.015）. The clinical model showed an AUC of 0.787 （95%CI 0.547-1.000） in the training group and 0.764 （95%CI 0.640-0.888） in the validation group. The ultrasound radiomics model showed an AUC of 0.806 （95%CI 0.605-1.000） in the training group and 0.806 （95%CI 0.711-0.901） in the validation group. The combined model significantly improved the AUC to 0.875 （95%CI 0.712-1.000） in the training group and 0.875（95%CI 0.792-0.959） in the validation group. DeLong test results showed that the performance of the combined model was superior to the clinical model （P = 0.015）. Decision curve analysis further confirmed the clinical utility of the combined model. ConclusionsThe combined model significantly improves the predictive ability for pCR after NAC in breast cancer patients through integrating miRNA-34a expression levels and ultrasound radiomics features. The potential biological associations were observed between ultrasound radiomics features and molecular markers, providing new tools and theoretical support for personalized treatment.